AI Agent Examples: Twenty Agents a 50-Person B2B Company Could Run

A 50-person B2B company could run around twenty useful AI agents across sales, marketing, client delivery, finance and leadership. Each takes on a frequent, defined job, reads the company's own knowledge, and stops for a person before anything goes out. The examples below use Halcyra, a fictional software company, to stay concrete.

What AI agents could a 50-person company run?

Around twenty, spread across every team: agents that research and draft for sales, produce and check content for marketing, prepare reports and handovers for delivery, reconcile and summarise for finance, and brief leaders. None of them sends, publishes or spends on its own. Each prepares work for a person to approve.

Adoption is still early. According to McKinsey's 2025 State of AI survey, 62% of organisations are at least experimenting with AI agents. According to BCG's 2025 AI at Work survey, only 13% of employees saw agents deeply integrated into their daily workflows. Most companies are choosing their first few, which is why concrete examples help.

Anthropic's engineering guidance is worth keeping in mind throughout: find the simplest solution possible, and add complexity only when needed. Several of these "agents" are really well-designed workflows with an AI step, which is exactly right.

Which AI agents help a sales team?

Five sales agents cover most of the groundwork: account research, pre-meeting briefs, post-call follow-ups, proposal first drafts and CRM notes from calls. They save salespeople hours each week while leaving every conversation, every judgement call and every sent message with a person.

1. Account researcher. Checks target accounts against the ideal customer profile and summarises news, hiring and funding.
2. Meeting briefer. Builds a one-page brief from CRM history, objections and recent account news before each call.
3. Follow-up drafter. Turns a call transcript into a follow-up that answers the objection raised, in the brand voice.
4. Proposal starter. Assembles a first draft from winning past proposals, approved claims and the client's brief.
5. Notes keeper. Summarises calls into CRM-ready notes for the salesperson to confirm.

Which AI agents help a marketing team?

Five marketing agents handle drafting, repurposing, checking, auditing and research: turning briefs into first drafts, reshaping long pieces for other channels, checking drafts against the voice and approved claims, auditing pages, and researching topics. A person approves everything that is published under the company's name.

1. Draft writer. Brief to first draft using product facts, customer quotes and the voice guide.
2. Repurposer. One article becomes a LinkedIn post, a newsletter section and social snippets.
3. Brand checker. Flags banned phrases, unsourced claims and off-brand punctuation in any draft.
4. Page auditor. Checks landing pages against an SEO and conversion checklist and lists fixes.
5. Topic researcher. Gathers sources and questions for a planned article, with every figure linked.

Which AI agents help client delivery?

Four delivery agents keep client work moving: weekly client reports, onboarding handovers from sales to delivery, meeting actions, and knowledge capture from client calls. They reduce dropped balls between teams and make every report consistent, whichever account manager happened to prepare it that week.

1. Report builder. Pulls the week's figures, compares them with targets and drafts commentary for an account manager to approve.
2. Handover writer. Turns the sales record into a delivery brief so the client does not repeat themselves.
3. Actions tracker. Extracts decisions and actions from meetings and drafts the recap email.
4. Knowledge sweeper. Files new facts from client calls into the shared knowledge, sending conflicts to a person.

Which AI agents help finance and operations?

Three finance and operations agents take on routine checks: matching invoices and flagging exceptions, summarising contracts against a checklist, and answering staff questions about internal policies. These back-office jobs are often where AI pays most, even though they get less attention.

1. Invoice checker. Matches invoices to purchase orders and flags exceptions for a person.
2. Contract summariser. Summarises new contracts against a checklist of terms to review.
3. Policy helper. Answers staff questions from the current policies, citing the source document.

Research backs the back-office emphasis. MIT NANDA's 2025 study, as reported by Fortune, found more than half of generative AI budgets go to sales and marketing, while the largest returns were found in back-office automation.

Which AI agents help leadership?

Three leadership agents turn information into decisions: a weekly business summary from the main systems, board pack first drafts, and competitor monitoring. They give leaders a consistent, sourced view of the business each week without someone spending days assembling it by hand.

1. Weekly summariser. One page on pipeline, delivery and cash, drawn from the core systems.
2. Board pack drafter. First draft of the board pack from agreed metrics and commentary.
3. Competitor watcher. Tracks competitors' public announcements, pricing pages and reviews, with sources.

Which AI agents should you build first?

Start with the two or three that save the most time on weekly jobs with easy-to-check output. For most B2B firms that means the follow-up drafter, the report builder and the meeting briefer. Prove them against a baseline, then add the next few. Twenty agents built at once is twenty half-finished agents.

The guide to building an AI agent covers the steps for each. Every one of these needs the same foundation: shared knowledge about the business, written methods and clear approval points.

Skein builds agents like these inside a Company OS, each reading a shared company brain; the agents page shows the full set with what each reads and produces. That suits a growing B2B company ready to systemise its frequent work. A very small team may need only two or three of the agents above.

Frequently asked questions

What are examples of AI agents in business?

Common examples include agents that research accounts before sales calls, draft follow-ups from call transcripts, build weekly client reports, turn briefs into first-draft content, check drafts against brand rules, match invoices to purchase orders and summarise contracts. Each prepares work for a person to approve.

How many AI agents does a business need?

There is no fixed number. A 50-person B2B company might eventually run around twenty, but should start with two or three on the jobs that waste most time. Prove each against a baseline before adding more, so every agent earns its place.

Which department benefits most from AI agents?

Sales and marketing get most of the attention, but back-office work often pays more. MIT NANDA's 2025 research, as reported by Fortune, found the largest returns in back-office automation even though most budgets went to sales and marketing tools. Look at finance and operations early.

What should an AI agent never do on its own?

An agent should not send messages to customers or prospects, publish content, change budgets or bids, make pricing commitments or bulk-edit records without a person approving. Those actions carry the company's name or money and are hard to undo once done.

Do AI agents need company data to work?

Yes, to be useful. An agent without the company's product facts, customer insight, voice and methods produces generic output. The best agents read a shared, maintained set of company knowledge before every task, and connect only to the systems their job requires.

Related reading